Today, the Alibaba Cloud Tongyi team officially released the new mathematical reasoning process reward model Qwen2.5-Math-PRM. This model offers two sizes, 72B and 7B, with performance significantly outperforming similar open-source process reward models, especially excelling in identifying reasoning errors. The 7B version of Qwen2.5-Math-PRM astonishingly surpasses the widely popular GPT-4o, marking an important step in Alibaba Cloud's research and development of reasoning models.

The Alibaba Qwen team recently published a paper titled 'Lessons Learned from the Development of Process Reward Models in Mathematical Reasoning' and introduced two new models in the Qwen2.5-Math-PRM series, featuring 7B and 72B parameters respectively. These models break through the limitations of the existing PRM framework in mathematical reasoning, significantly improving the accuracy and generalization ability of reasoning models through innovative techniques. Mathematical reasoning has long been a major challenge for large language models (LLMs), especially regarding errors in intermediate reasoning steps.

Recently, Alibaba Damo Academy launched a multi-modal large language model named Valley 2, designed for e-commerce scenarios. It aims to enhance performance across various fields and expand the application boundaries of e-commerce and short video scenarios through a scalable vision-language architecture. Valley 2 utilizes Qwen 2.5 as its LLM backbone, paired with the SigLIP-384 visual encoder, incorporating MLP layers and convolution for efficient feature transformation.

The Qwen team recently announced the open-source release of their latest multimodal reasoning model, QVQ, marking an important step forward in the capabilities of artificial intelligence in visual understanding and complex problem-solving. This model is based on Qwen2-VL-72B and aims to enhance AI reasoning capabilities by combining language and visual information. In the MMMU evaluation, QVQ achieved a high score of 70.3, demonstrating significant performance improvements over Qwen2-VL-72B-Instruct in several math-related benchmark tests.

The Alibaba Cloud Bairen Model Service Platform recently launched the 'Real-Time Audio and Video Interaction' feature, enabling users to easily build multi-modal AI applications without the need for programming knowledge. This new feature allows users to quickly integrate AI models into Web, iOS, and Android applications and share them with others. Users can create intelligent agent applications through simple steps: first, create a new intelligent agent application, then select and configure the required text, speech, or visual understanding models on the Alibaba Cloud Bairen platform. The platform offers over 200 large models, including

Recently, Alibaba announced the open sourcing of its latest image generation model, Qwen2vl-Flux. This model not only features a variety of functionalities such as editing, fusion, and mixing, but it can also generate highly similar new images based on user-input images or text. Qwen2vl-Flux provides powerful image transformation capabilities. Users simply need to input an image without any text prompts, and the model can generate multiple similar images based on the original. For example, if a user uploads a portrait photo, the model can generate multiple angles.

The Aliyun Bailing platform recently announced the launch of the Qwen2.5-Turbo million long text model, developed by the Tongyi Qianwen team. The Qwen2.5-Turbo model supports processing ultra-long contexts of up to 1 million tokens, equivalent to 1 million English words or 1.5 million Chinese characters.

On November 15, Alibaba Tongyi Lab launched the Code Mode, a feature that enables users to generate various applications through simple everyday language instructions, including mini games, data charts, websites, and resumes. Users can experience this new interactive method by visiting the Tongyi web version (tongyi.ai) and clicking on 'Code Mode' in the input box.

The Tongyi Qianwen team recently announced the open sourcing of their latest Qwen2.5-Coder full series, aimed at promoting the development of Open Code LLMs. Qwen2.5-Coder has garnered attention for its power, diversity, and practicality. The Qwen2.5-Coder-32B-Instruct model has achieved SOTA levels in coding capabilities, comparable to GPT-4o, demonstrating comprehensive abilities.

Recently, an open source multimodal AI model named Molmo has drawn widespread attention in the industry. This AI system, based on Qwen2-72B and leveraging OpenAI's CLIP as the visual processing engine, is challenging the dominance of traditional commercial models with its outstanding performance and innovative features. Molmo's standout characteristic is its efficient performance. Despite its relatively small size, it can compete with competitors that are ten times larger in processing capability. This 'small but exquisite' design philosophy not only enhances the model's

The 2024 Yunqi Conference opened grandly in Hangzhou, with Alibaba Group CEO and Alibaba Cloud Intelligence Group Chairman and CEO Wu Yongming delivering a keynote speech. During the conference, Alibaba announced a series of groundbreaking releases, including the launch of the next-generation open-source model Qwen2.5, which outperforms Llama405B and firmly holds the position of the 'strongest open-source large model'.